Evaluation of articulatory similarity using formant and fundamental frequencies during perceptual assimilation of English schwa by native speakers of Japanese

The main hypothesis behind the perceptual assimilation model (PAM) is that listeners perceive non-native speech sounds based on articulatory similarities between the non-native speech sounds and native speech sounds. When non-native and native sounds are substantially similar, non-native sounds are perceived as equivalent to native ones. Former research shows that the schwa vowel is perceptually assimilated to Japanese /a/ almost exclusively. In this study, we investigate acoustical cues that Japanese listeners rely on to assess the articulatory similarities between the English schwa and the Japanese vowels during the perceptual assimilation. Traditionally, the first two formants (F1 and F2) are considered to be effective for assimilation judgements; however, former investigations imply that these dimensions may be insufficient. In this study, we compared the schwa and the five Japanese vowels in the dimensions of vowel openness and backness using additional information, i.e., the third formant (F3) and fundamental frequencies (F0). The results of the analyses suggest that Japanese listeners use F1, F2, F3 and F0 information to assess articulatory similarities between the schwa and the Japanese vowels during perceptual assimilation.